Sign in to view source links and access this dataset
Description
A predicted spatial transcriptomics atlas for The Cancer Genome Atlas (TCGA) histopathology slides, generated using the DeepSpot-M model. The atlas covers both fresh-frozen (FF) and formalin-fixed paraffin-embedded (FFPE) tissue samples. It was created by researchers including Kalin Nonchev, Sebastian Dawo, Karina Silina, Viktor Hendrik Koelzer, and Gunnar Rätsch, with the associated code available on GitHub.
Use Cases
Train or validate spatial gene expression prediction models based on H&E slide images.
Analyze spatial patterns of gene expression in TCGA cancer samples.
Correlate predicted transcriptomic profiles with clinical outcomes from TCGA.
Benchmark new computational pathology tools against the DeepSpot-M predictions.
Strengths
Covers both FF and FFPE tissue preparation types from the TCGA repository.
Generated using a specific, documented method (DeepSpot-M) with code publicly available.
Last updated on 2026-06-20, indicating recent maintenance.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Data is predicted via a model rather than directly measured, which may introduce artifacts.
Provenance
Source
The Cancer Genome Atlas (TCGA) H&E slides.
Collection Method
Predicted using the DeepSpot-M model.
Freshness
Last updated 2026-06-20 14:32:49; freshness should be verified.
Files are stored as gzip-compressed archives; specific formats and structure require inspection of the full dataset page.